JP3455752B2 - Age estimation device and age display control method - Google Patents

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION The present invention relates to an age estimation device and year.
The present invention relates to an age display control method , more specifically, an age estimation device and an age table capable of estimating age by inference from a face or a face image.
Indication control method .

[0002]

2. Description of the Related Art Generally, human beings have the ability to infer the age of a person without looking at the face of others, but from the viewpoint of managing their own health, the age of their own face is There are times when you want to objectively check whether the age is close to the average person's face or is extremely distant. Also, in the criminal investigation, in order to make the arrest of the criminal effective, it is sometimes desired to estimate the age of the criminal from the montage photograph created by the criminal witness.

[0003]

[Problems to be Solved by the Invention]
There is nothing that can satisfy such a demand. Therefore, for example, it has not been possible to quickly and easily know the corresponding objective age from the face image. Moreover, it was not possible to easily know the result of comparison with the average age.

Therefore, according to the present invention, the age can be objectively and quickly estimated from the created face or face image, and the actual age that is input and displayed and the face or face can be estimated.
Display the size with the estimated age, estimated based on the image
You can compare and look at the inside face or face image,
Whether the face or face image being displayed looks younger than the actual age,
Promptly and surely confirming the judgment of whether to look old
It is an object of the present invention to provide an age estimation device and an age display control method capable of controlling the age .

[0005]

To achieve the above object, the age estimating device according to the present invention has the following configuration. That is, the invention according to claim 1 includes a face image creating means for creating a face image by combining a plurality of part images, and a face feature detecting means for detecting a feature of the face image created by the face image creating means. , Performing fuzzy inference according to a predetermined fuzzy rule using the feature detected by the face feature detecting means as an input parameter, and estimating the age of the face image created by the face image creating means based on the result of the fuzzy inference. Age estimation means and face image creation
Enter the actual age corresponding to the face image created by
The age input method to be used and the age input method
Actual age, estimated by the age estimation means
A certain age and created by the face image creation means
Display to control the displayed face image on the same screen
And a control means .

According to the second aspect of the present invention, a face feature detecting means for detecting face features and face feature data representing the face features detected by the face feature detecting means are used as input data for a predetermined fuzzy operation. The fuzzy inference is performed according to a rule, and the age estimating means for estimating the age of the face based on the result of the fuzzy inference, and the face feature detecting means.
Age input to enter the actual age corresponding to the detected face
Means and the actual year entered by this age input means
Age, the estimated age estimated by the age estimating means,
The face detected by the face feature detection means
Display control means for controlling the same screen to be displayed .

According to a third aspect of the present invention, there is provided a first display control means for displaying the stored face, and a face feature for detecting the feature of the face displayed under the control of the first display control means. The detecting means, the age estimating means for estimating the age of the face based on the face feature detected by the face feature detecting means, and the control of the first display control means
After displaying, enter the actual age for the face
Age input means and the actual input by this age input means
Age and estimated age estimated by the age estimating means
So that and are displayed on the same screen as the displayed face.
And a second display control means for controlling .

[0008]

According to the first aspect of the invention, a face image is created by combining a plurality of part images, the features of the created face image are detected, and then the detected features are used as input parameters to determine a predetermined value. It is possible to perform fuzzy inference according to a fuzzy rule and estimate the age of the created face image based on the result of the fuzzy inference. Meanwhile, before
Enter the actual age corresponding to the created face image
And the actual age entered, the estimated
Display the age and the created face image on the same screen.
Can be shown . According to the invention of claim 2, the facial features are detected, fuzzy inference is performed according to a predetermined fuzzy rule using the facial feature data representing the detected facial features as input data, and based on the result of the fuzzy inference. The age of the face can be estimated. Meanwhile, the detection
Enter the actual age corresponding to the face
Actual age, estimated age, and
And displaying the detected face on the same screen
You can According to the invention of claim 3, when the stored face is displayed and then the feature of the displayed face is detected, the age of the face is estimated based on the detected face feature. be able to. On the other hand, the face was displayed
After that, if you enter the actual age corresponding to the face, this
Forced actual age and said estimated estimated age
It can be displayed on the same screen as the displayed face.
It

Therefore , according to the present invention, it is possible to quickly and easily estimate the objective age from the face or the face image without requiring any special skill. Not only that, but the actual age entered and displayed and the face or face
Display the size with the estimated age, estimated based on the image
You can compare and look at the inside face or face image,
Whether the face or face image being displayed looks younger than the actual age,
Promptly and surely confirming the judgment of whether to look old
You can

[0010]

Embodiments of the present invention will be described below with reference to the drawings. FIG. 1 is a block diagram showing an embodiment of an age estimating device according to the present invention. In FIG. 1, the age estimating device is roughly divided into a CPU 1, a characteristic switch 2, a cursor switch 3, a display switch 4, a position correcting switch 5, an age estimating switch 6, an age input key 7, a program ROM 8,
Work RAM 9, parts screen ROM 10, VRAM 1
1, a table ROM 12 and a display unit 13.

The CPU 1 controls the entire apparatus, and includes a characteristic switch 2, a cursor switch 3, a display switch 4,
When the position correction switch 5, the age estimation switch 6, and the age input key 7 are used to specify the face parts, features, etc., according to the operation information, according to the program stored in the program ROM 8, the age data is displayed. A face image (for example, one's face or the face of another person) is read out by reading out the face feature data from the table ROM 12, and reading out and combining the corresponding part patterns for each part from the part screen ROM 10 according to the face feature data. Perform the process of creating. Also, the feature data corresponding to the age of the created face image is detected, fuzzy inference is performed according to a predetermined fuzzy rule using the detected data as an input parameter, and the age of the face image created based on the inference result is calculated. Perform an estimation process. The CPU 1 has functions as a face image creating means, a feature data detecting means, a face feature detecting means, an actual age designation determining means, and an age estimating means.

The feature switch 2 is a switch for designating the features of the face, and sequentially designates each part (for example, hairstyle, face outline, eyebrows, eyes, nose, mouth, etc.) that constitutes the face. In this embodiment, six parts are prepared as each part that constitutes the face. For the face image displayed on the display unit 13, the cursor switch 3 is provided for each part (eg,
Hair, eyebrows, eyes, nose, mouth, etc.). Specifically, the part is selected by moving the cursor position on the screen (for example, represented by a blinking state). A mouse or the like may be used instead of the cursor switch 3. The use of the mouse is the same for other switches.

The display switch 4 is operated when the face image is displayed on the display unit 13 for correction, or when the created face image is in a desired state or set as a finished screen. The position correction switch 5 is operated when correcting the position of each part (especially, eyebrows) forming a face, and is composed of, for example, an up and down push button type switch. The age estimation switch 6 is operated when estimating the age of the created face image. The age input key 7 is used for inputting and specifying data relating to age, and as data relating to age, there are actual age, date of birth, year before, and year after. As the age input key 7, for example, a ten-key is used so that the age or date of birth can be easily entered. The age input by the age input key 7 is displayed on the display unit 13 as shown in FIG.

The program ROM 8 stores a control program for the CPU 1, the contents of which are shown in the respective figures described later. The work RAM 9 is used as a work area in the control performed by the CPU 1. The parts screen ROM 10 stores a parts pattern for each part representing a face image, and constitutes a parts pattern storage means. Here, as the face part types (N) stored in the parts screen ROM 10, there are various face parts, and a plurality of types are prepared, for example, hairstyle, face outline, eyes, eyebrows, nose, mouth. ing. In addition to this, parts types (N) such as shoulders, chest, torso, and feet may be prepared.

The part pattern is a modified mode of each part (particularly, a part pattern having a shape according to age)
For example, a plurality of types of male hairstyles corresponding to various types are provided in advance, and the parts screen ROM 1
It is stored in 0. In this case, various parts patterns that form the face image are prepared for each part such as the contour of the face, eyes, and eyebrows, and according to age. Also,
For each part pattern, variations thereof are prepared and stored in advance. For example, there are 16 types of contour patterns as variations of facial contour patterns. These are provided according to the actual situation in which various face contours exist.

The VRAM 11 is used when a face image is created.
The created image is stored on a screen-by-screen basis.
For example, a semiconductor memory is used as 11. The table ROM 12 stores face characteristic data according to age and constitutes a face characteristic data storage means. The face feature data according to age includes face aspect ratio, eye height with respect to the whole face, eye size, number of wrinkles, receding degree of hair. The facial feature data stored in the table ROM 12 will be described in detail. For example, the vertical axis represents the facial features, such as wrinkles around the eyes, wrinkles on the forehead, and wrinkles on the cheeks.
The facial feature data is prepared in advance in such a manner that the features up to 100 are arranged by dividing the ages from 30 to 5 on the horizontal axis in sequence on the horizontal axis. In this case, for example, paying attention to the wrinkles on the outer corners of the eyes, there is no wrinkle on the outer corners of the eyes at the age of 30, but the facial features (in this case, the wrinkles on the outer corners of the eyes) are stored in advance so that the wrinkles on the outer corners of the eyes increase as the age increases. Has been done. Further, the table ROM 12 stores table data representing a membership function used in fuzzy inference described later.

A display unit (display means) 13 displays a face image created by the CPU 1, an estimated age, etc., and selects each part during creation while exchanging data with the VRAM 11. The displayed face image is displayed, or the completed face image and estimated age are displayed. The display unit 13 has, for example, a TV display for displaying a face image and the like. It should be noted that the portion for displaying images and the like is not limited to the TV display, and may be, for example, a dedicated monitor device or a CRT, and may be an LC
An image may be displayed by a liquid crystal such as D. Alternatively, one that is also used for other purposes may be used.

Here, the fuzzy rule of the age estimation process will be described. 2A and 2B are membership functions of the antecedent part, of which FIG. 2A is a membership function relating to the aspect ratio of the face, and FIG. 2B is a membership function relating to the position of the eyes. . Further, FIG. 2C is a fuzzy output in the consequent part, which is a membership function regarding age.

The meaning of the label in each membership function is as follows. FS: "Face aspect ratio is small" FM: "Face aspect ratio is medium" FL: "Face aspect ratio is large" EL: "Eye position is low" EM: "Eye position is medium" EH: "Eye position is high" YY: "Age is young" YM: "Age is middle age" YO: "Age is high" In addition, "Age is middle age" is a reference age (for example, , 30
The age may be appropriately set according to the target person group.

The fuzzy rules are so-called IF, THE
It is expressed in the form N (if, if). In this embodiment, the following three rules are adopted. Rule A: IF Face Aspect Ratio = FL AND Eye Position =
EL THEN Fuzzy output = YY (younger age) The fuzzy rule A means "if the face aspect ratio is large and the eye position is low, the age is young." Rule B: IF Face Aspect Ratio = FM AND Eye Position =
EM THEN Fuzzy output = YM (age is middle age) The fuzzy rule B means "if the face aspect ratio is medium and the eye position is medium, the age is middle age". Is. Rule C: IF Face Aspect Ratio = FS AND Eye Position =
EH THEN Fuzzy output = YO (older age) The fuzzy rule C means "if the face aspect ratio is small and the eye position is high, the age is high".

Next, the operation will be described. FIG. 3 is a flowchart showing a main program of face image creation / age estimation processing. When this program starts, first, in step S10, initialization is performed. In the initial setting, various registers, the work RAM 9, the VRAM 11 are cleared, the subroutine is initialized, the flags are reset, and the like. Next, in step S12, the pointer M is cleared to [0]. The pointer M is used as a designation when inputting information such as facial features and age, and its value is C
It is stored in the corresponding register in PU1.

Next, in step S14, the contents of the VRAM 11 are displayed on the display unit 13. As a result, for example, a face image created by combining the required part patterns for each part in the process of creation, a completed face image, and a fuzzy inference result, and the age estimated based on the estimation result. The face image at that time is displayed on the display unit 13 (see FIG. 16). After that, this step S1
4, the necessary processing is performed based on the interrupt signal from each switch. That is, the processing for selecting each part is executed by the interrupt routine shown below.

FIG. 4 is a flowchart showing the feature switch interrupt routine. When the characteristic switch 2 is operated, the routine proceeds to this characteristic switch interrupt routine. When the process shifts to this interrupt routine, first, step S20.
The pointer M is incremented by [1]. Next, in step S22, it is determined whether or not the pointer M is equal to [9]. Here, the pointer M is compared with [9] in order to determine whether or not the value is [9], which is a value exceeding eight, because there are eight pieces of facial feature data representing facial characteristics. is there.

If the pointer M is not equal to [9], the process proceeds to step S26, where (M + O in the work RAM 9).
The contents of the screen data are transferred to the VRAM 11 using the data of the address FFSET1) as the start address. For example,
When M = 1, the data at the address (1 + OFFSET1) becomes the start address and the contents of the screen data in the work RAM 9 are transferred to the VRAM 11.

Here, as shown in FIG. 5, the work RAM 9 mainly stores OFFSE from the address (OFFSET1 + 1).
Various necessary data are temporarily stored until the address T4. For example, in the area having the address (OFFSET1 + 1) as address, data ADD1 designated by the pointer M = 1 is stored, and this data corresponds to the address of each part stored in the parts screen ROM 10. Similarly, in the area whose address is (OFFSET1 + 2), the data ADD2 designated by the pointer M = 2 is stored.
The area whose address is T1 + 7) is the pointer M
Data data ADD7 designated by = 7 is stored.

The other areas of the work RAM 9 will be described. (OFFSET2 + 1) address, (OFFSET2 +)
2) Address ... Stores data corresponding to the cursor position. (OFFSET3 + 1) address, (OFFSE
The address T3 + 2) ... Stores data representing facial features. Specifically, the area whose address is (OFFSET3 + 1) stores hairstyle data, and the most significant bit (portion indicated by a) of the data stores a gender parameter for distinguishing between male and female. It The gender parameter is a 1-bit unit, and when it is [1], it represents a male, and when it is [0], it represents a female.

Further, in the area whose address is the address (OFFSET3 + 1), data on the hairstyle is stored in the upper bit area (upper bit group) b, and the lower bit area (lower bit group) in the same area is stored. Characteristic data according to age regarding the hairstyle is stored in c. (OF
The area whose address is FSET3 + 2) stores data relating to the contour of the face, and the sex parameter is stored in the most significant bit a of the data. (OFF
Data regarding the contour of the face is stored in the upper bit group b in the area whose address is SET3 + 2), and characteristic data regarding the contour of the face is stored in the lower bit group c in the area. (OFFSET3 +
3) The area whose address is the address stores data relating to the eyes, and the sex parameter is stored in the most significant bit a of the data. In the area whose address is (OFFSET3 + 3), the upper bit group b stores the eye data, and the lower bit group c in the same area stores the eye-related feature data.

Thereafter, similarly (OFFSET3 + 4)
Areas whose addresses are addresses sequentially store data about eyebrows (eg, nose, mouth, etc.)
Is stored. Among the characteristic parts of the face, the parts that change significantly according to age are the hairstyle, outline, and eye
Therefore, as the characteristic data according to age, the feature data corresponding to these three types (hairstyle, contour, eyes) changed according to age is stored as characteristic data. O
Data related to age is stored in the FFSET address 4.

Now, for example, as described above, the pointer M =
When it is 1, the data of the address (1 + OFFSET1) becomes the start address and the content of the screen data of the work RAM 9 is transferred to the VRAM 11. At this time, the work R
The data at the address (1 + OFFSET1) in AM9 is ADD1 and its content is one of the screens for each part stored in the parts screen ROM 10 as shown in FIG. In FIG. 10, the screen data corresponding to the address of ADD1 is a screen for selecting the sex, and therefore when M = 1, the screen data corresponding to ADD1 is [01]:
A screen prompting to specify either male or [02]: female appears.

Similarly, in FIG. 6, the screen data corresponding to the address of ADD2 is a screen for selecting a hairstyle. For example, [01] is seventy-three divisions with a rich hair amount, and [02] is a thin hair amount. Seventy-three divisions, and so on. A screen for prompting the selection of a hairstyle is displayed corresponding to this ADD2. The screen data corresponding to the address of ADD3 is a screen for selecting the contour of the face, and for example, [01] is a round shape, [02] is a square shape, and so on. A screen for prompting the selection of the contour of the face appears corresponding to this ADD3. The screen data corresponding to the address of ADD4 is a screen for selecting an eye, for example, [01]
Has double eyelids with round eyes, [02] has oval eyes, ...
It is supposed to be. Corresponding to this ADD4, a screen prompting the selection of the shape of the eyes appears.

The screen data corresponding to the address of the ADD 5 is a screen for selecting eyebrows. For example, [01] has crescent-shaped eyebrows, [02] has eyebrow-shaped, ... Corresponding to this ADD5, a screen prompting the selection of the eyebrow shape appears. The screen data corresponding to the address of the ADD 6 is a screen for selecting the nose. For example, [01] has a large and high nose, [02] has a small and low nose, and so on. This ADD
A screen for prompting the selection of the nose corresponding to 6 appears. A
The screen data corresponding to the address of the DD 7 is a screen for selecting a mouth, and is, for example, [01], [02], ... Corresponding to this ADD7, a screen prompting the selection of the mouth appears. In addition to this, ADD
The screen data corresponding to the addresses after 8 is a screen for requesting the input of age, and a screen for prompting the input of the age from the age input key 7 appears correspondingly. Further, there is an area for storing a part pattern for each part as screen data corresponding to addresses after ADD8. Specifically, various part patterns such as a hairstyle part pattern, a face contour part pattern, an eye part pattern, ... Are stored.

Returning to the explanation of FIG. 4 again, after the processing of step S26, the process returns. Then, when the next feature switch interrupt routine is reached, similar processing is repeated.
At this time, first, the pointer M is incremented in step S20, and when the pointer M becomes equal to [9] in step S22, the process proceeds to step S24 and the pointer M is set to [1].
To step S26. Thus, every time the characteristic switch 2 is operated, the pointer M is moved to [1].
The pointer M is incremented by one, and when the pointer M becomes equal to [9], the pointer M is returned to [1] again. Therefore, the work RA corresponding to the pointer M = 1
The contents of the screen data are transferred to the VRAM 11 by using the data of the address (M + OFFSET1) of M9 as the start address, and thereafter, every time the characteristic switch 2 is operated, the pointer M is incremented by [1] (M + OFFSE).
The contents of the screen data at the address T1) are transferred to the VRAM 11 and displayed on the display unit 13.

Each time the characteristic switch 2 is operated, the contents of the screen data change according to the data corresponding to ADD1 to ADD7 in FIG. That is, each time the characteristic switch 2 is operated, each time the characteristic switch 2 is operated, the characteristic switch interruption routine is executed, and a screen corresponding to the data corresponding to ADD1 to ADD7 in FIG. 6 appears. The operator selects facial features such as gender, hairstyle, eyes, eyebrows, ... While watching these screens.

FIG. 7 is a flowchart showing the cursor switch interrupt routine. When the cursor switch 3 is operated, the process proceeds to this cursor switch interrupt routine. In this interrupt routine, it is first determined in step S80 whether face image data is being displayed (for example, a display flag is used). This is because it is determined whether or not the screen of the face image is displayed on the display unit 13, and if the face image is displayed, the cursor does not appear. Therefore, if the image data is being displayed, this routine returns.

On the other hand, when the face image data is not being displayed, the operation proceeds to the subsequent step S82, where (M + O of the work RAM 9 is displayed.
FFSET2) Change the data of the address according to the cursor position. Since the OFFSET address 2 stores data corresponding to the cursor position, for example, when M = 2, the data of the (OFFSET2 + 1) address for the hairstyle is changed according to the cursor position. Further, when M = 3, the data of the address OFFSET2 + 2 regarding the contour is changed according to the cursor position. For example, when the cursor switch 3 is operated to select a part pattern, one of the variations of various part patterns related to the contour is selected with the aerosol.

Next, in step S84, the display of the cursor position is changed. As a result, when the cursor switch 3 is operated and moved, the display position is changed and displayed on the screen. Next, in step S86, it is determined whether or not the pointer M is [1]. That is, it is determined whether or not the screen is a gender determination screen. When M = [1], the process proceeds to step S88, and it is determined whether or not the data at the (OFFSET2 + M) address represents a man. If it represents a man, it is first reset to N = 1 in step S90. N is a gender pointer. This is because the gender pointer N is sequentially incremented from the minimum value, that is, the area corresponding to the (OFFSET2 + 1) address. (O
The address FFSET4 + M) stores data representing facial features, and the most significant bit MSB (portion indicated by a) of the data stores a gender parameter for distinguishing between male and female. Gender parameter is 1 bit unit,
"1" represents a male, and "0" represents a female.

Then, the process proceeds to step S92 (OFF
Set the MSB of the data at address SET3 + N) to "1". At this time, since N = 1, first (OFFSE
The MSB of the data at address T4 + 1) will be set to "1". Next, in step S94, the gender pointer N is incremented. As a result, N = 2.
Next, in step S96, it is determined whether N = 7. Judging N = 7 is (OFFSET3 + N)
This is to set the MSB to "1" up to the maximum address. Since N = 7 this time, step S92
Return to and repeat the same processing. Therefore, next time N
= 2, the MSB of the data at the address (OFFSET3 + 2) is set to "1". And so on
(OFFSET3 + 3) address ... (OFFSET3
The MSB of each data at address +6) is set to "1".

When N = 7 in step S96, the current cursor switch interrupt routine is terminated and the process returns to the main program. On the other hand, step S
If the pointer M is not [1] at 86, step S1
Jump to 06 and change the contents of the (OFFSET2 + M) address to the M of the data of the {(OFFSET3 + (M-1)} address.
Transfer to upper bit group b excluding SB. As a result, the data content at the position designated by the cursor switch 3 is transferred to the area for storing the facial feature data. After step S106, the cursor switch interrupt routine is terminated and the process returns to the main program. If NO in step S88, that is, if the data at the address (OFFSET2 + M) does not represent a man but a woman, the process proceeds to step S98 and the gender pointer N is returned to N = 1. This is because the gender pointer N is sequentially incremented from the minimum value, that is, the area corresponding to the (OFFSET2 + 1) address.

Then, in step S100 (OFFSE
The MSB of the data at address T3 + N) is set to "0". At this time, since N = 1, first (OFFSET
The MSB of the data at address 3 + 1) will be set to "0". Next, in step S102, the gender pointer N is incremented. As a result, N = 2.
Next, in step S104, it is determined whether N = 7. It is (OFFSET3 +
This is to set the MSB to "0" up to the maximum address of N). This time N = 7, so step S
Returning to 100, the same processing is repeated. Therefore, the next time N = 2, the MSB of the data at the address (OFFSET3 + 2) is set to "0". Similarly, in the same way, (OFFSET3 + 3), ... (OFFSE
The MSB of each data at address T3 + 6) is set to "0". Then, in step S104, N = 7
Then, the cursor switch interrupt routine this time is terminated and the process returns to the main program.

FIG. 8 is a flow chart showing the display switch interrupt routine. When the display switch 4 is operated, the process proceeds to this display switch interrupt routine. When the process shifts to this interrupt routine, first, step S15.
The display flag is inverted with 0. Therefore, the display flag is inverted every time the display switch 4 is operated. For example, the display flag is “0” at first, and when operated, the display flag becomes “1”, and thereafter, “0” and “1” are set for each operation.
repeat. In this way, it is determined whether or not the display switch 4 has been operated in the previous routine.

Next, in step S152, it is determined whether or not the display flag is "1". The display flag being “1” is, for example, the case of the first routine. In this case, the process branches to step S154 (OFFSET1 +
M) The screen data having the content stored in the address as the start address is transferred to the VRAM 9. At this time, M = 1
If so, (1 + OFFSET in the work RAM 9
1) The address data ADD1 is read. AD
The screen data corresponding to the address of D1 is a screen for selecting the gender, and therefore, a screen prompting the selection of the gender is displayed on the display unit 13 at first corresponding to ADD1. According to this screen, the operator operates the cursor switch 3 to specify either [01]: male or [02]: female. After step S154, the process returns to the main program.

On the other hand, after the completion of the first routine, the display flag is inverted every time the display switch 4 is operated, so that the next time the determination result in step S152 is YES, the process proceeds to step S156. In step S156, the pointer M is returned to [1]. Pointer M = 1 is gender, M =
2 corresponds to a hairstyle, M = 3 corresponds to a contour, M = 4 corresponds to an eye, and so on. Therefore, in order to return the pointer M to [1] in step S156, the gender is first determined, Then
This is for sequentially selecting and displaying characteristic data (for example, a part pattern) according to age from the hairstyle.

Then, in step S158 (M + OFF
The screen data having the content stored in the address SET3) as the start address is transferred to the VRAM 11. At this time, the data of the address (OFFSET3 + M) in the work RAM 9 is an area for storing the facial feature data.
Therefore, a face image corresponding to the face feature data is displayed on the display unit 1.
It is displayed in 3.

Next, in step S160, the pointer M is incremented, and in step S162, it is determined whether or not the pointer M is [7]. This is to determine whether or not the selection of all the six feature data of the hairstyle, the contour, and the eyes has been completed as the feature data of the face. M = [7]
If not, the process returns to step S158 and the same loop is repeated. In this way, the above processing is repeated in the range of the pointer M of [1] to [6], and M = M in step S162.
When [7] is reached, this routine is terminated and the process returns to the main program.

FIG. 9 is a flow chart showing the position correction switch interruption routine. When the position correction switch 5 is operated, the process proceeds to this position correction switch interrupt routine. When this interrupt routine is entered, it is first determined in step S200 whether the display flag is [1]. The display flag is inverted when the display switch 4 is operated as described above.

When the display flag is not [1] in step S200, it is determined that the face image is not displayed on the display unit 13, the routine of this time is ended, and the process returns to the main program. On the other hand, if the display flag is [1], it is determined that the face image is displayed, and the process proceeds to the subsequent step S202. In step S202, it is determined whether or not the operation result of the position correction switch 5 is to raise the position of the part. The position correction switch 5 is operated when correcting the positions of the parts (especially, eyebrows) that form the face, and is composed of push-button type up and down switches. Therefore, here, it is determined whether or not the up switch is pressed. Determine whether. If the up switch has not been pressed, it is determined that the down switch has been pressed and step S
Proceed to 204.

In step S204, (3 + OFFSET
3) The screen data is sequentially read using the content stored in the address as an address, the y coordinate of the screen data read in step S206 is converted into a value decremented by [1], and the value is transferred to the VRAM 11. As a result, the position of the eyes (in particular, the y coordinate position) changes so as to be lowered by the operation of the position correction switch 5, and the display unit 13 displays the face image in which the position of the eyes is corrected downward. . Generally, as the age increases, the eye position tends to increase, and when the eye position is corrected, the accuracy of age estimation improves. After step S206, the routine of this time is ended and the process returns to the main program.

On the other hand, when the up switch is pressed in step S202, the process proceeds to step S208 and (3 + O
The screen data is sequentially read by using the content stored in the address FFSET3) as an address, the y coordinate of the screen data read in step S210 is converted to a value incremented by [1], and the value is transferred to the VRAM 11. As a result, the position of the eyes (in particular, the y-coordinate position) changes so as to go up, and the face image in which the arrangement of the eyes has been corrected upward is displayed on the display unit 13. After step S210,
The routine this time is ended and the process returns to the main program.

FIG. 10 is a flowchart showing an age estimation switch interruption routine. When the age estimation switch 6 is operated, the process proceeds to this age estimation switch interrupt routine. When this interrupt routine is entered, it is first determined in step S300 whether the display flag is "1". When the display flag is not [1] in step S300, it is determined that the face image is not displayed, the routine of this time is ended, and the process returns to the main program.

On the other hand, if the display flag is [1], it is determined that the face image is displayed, and the processing from step S302 is executed. First, in step S302 (OFF
The contents stored in the address (SET3 + 2) (that is, the contour of the face) are read out, the aspect ratio of the face is calculated based on the data read out in step S304, and the calculated value is calculated by the CPU.
Store in the A register in 1. This is to obtain the aspect ratio of the created face image from the contour data of the face image, and has a reference aspect ratio in advance. This is because the age is estimated by fuzzy reasoning by detecting the aspect ratio of the face. Therefore, the content stored in the A register becomes one of the input parameters in fuzzy inference. In addition, the aspect ratio of the face image is calculated by measuring the length of the face image in the vertical direction (direction connecting the top of the head and the bottom of the chin) and the length in the horizontal direction (left and right direction of the face) from the face image. You may ask.

Then, in step S306 (OFFSE
The content (that is, the eye position) stored in the address T3 + 3) is read, and the eye position data is obtained based on the data read in step S308. Note that the position may be obtained by actually measuring the eye position from the face image. Next, the position data obtained in step S310 is corrected by the coordinate position (y coordinate position) corrected by the position correction switch 5, and the corrected value is stored in the B register in the CPU 1. In this case, the eye position of the created face image is obtained from the eye position data of the face image, and if the position correction switch 5 is operated at this time, the eye position is finally obtained by taking the correction value into consideration. It is a thing. If the position correction switch 5 is not operated, the correction value is [0] and nothing is added. This is because the age is estimated by fuzzy reasoning by detecting the position of the eyes. Therefore, the content stored in the B register becomes one of the input parameters in fuzzy inference.

Since the two input parameters (face aspect ratio and eye position) in fuzzy inference have been obtained as described above, fuzzy inference processing is performed next. First, in step 312, the aforementioned fuzzy rule A is calculated based on the aspect ratio of the face and the position of the eyes. Similarly, in steps S314 and S316, the aforementioned fuzzy rule B and fuzzy rule C are calculated based on the aspect ratio of the face and the position of the eyes. Detailed processing contents of these fuzzy rules A, B, and C will be described in a subroutine described later. Thereby, the processing of the antecedent part in the fuzzy inference is performed, and the goodness of fit for each membership function is obtained.

Next, in step S318, the process of taking the maximum value of the data obtained by the rule calculation (that is, MAX
Along with the calculation = OR processing, the center of gravity of this data is calculated in step S320 to defuzzify the data. Thus, the consequent part of the fuzzy inference is processed to obtain the estimated age j of the created face image. Detailed processing contents of the maximum value calculation and the center of gravity calculation will be described in a subroutine described later. Next, the estimated age j obtained in step S320 is transferred to the VRAM 11 and displayed on the display unit 13. As a result, the age can be objectively and quickly estimated from the created face image, and the result of comparison with the average age can be known.

FIG. 11 is a flow chart showing a subroutine of the fuzzy rule A arithmetic processing. First, in step S350, data (antecedent part data) Da is obtained from the FL table in the table ROM 12 in which the membership function "the aspect ratio is large" is stored by using the content of the A register (face aspect ratio) as an address. read out. The data in this case corresponds to the goodness of fit (grade) of the membership function FL shown in FIG. 2A, and has a value in the range of 0 to 1. That is, the data Da is the matching degree indicated by the rule A with respect to the input value of the aspect ratio of the face. Data D
a is stored in a register of the same name in the CPU 1, for example,
This also applies to data Db described later.

Similarly, in step S352, the contents of the B register (eye position) are used as addresses, and the table ROM 1
The data (the antecedent data) Db is read from the EL table in which the membership function "the eye position is low" in 2 is stored. The data in this case is shown in Figure 2.
It corresponds to the goodness of fit of the membership function EL shown in (b), and is a value in the range of 0 to 1. That is, the data Db is the matching degree indicated by the rule A for the input value of the eye position. Next, in step S354, it is determined whether or not the data Da is greater than or equal to the data Db. If Da ≧ Db, the process proceeds to step S356, and the smaller value is adopted as the data D. In this case, D = Db.

This process is to take the logical product of the respective data and is equivalent to the so-called "take MIN". This is to obtain a state that satisfies at least both of two input values by "taking MIN". As a result, the smallest one of the data Da and the data Db is adopted as the judgment indicated by the rule A in this routine. On the other hand, when Da <Db, the process proceeds to step S358, and D = Da is adopted. in this way,
For rule A, two input values at that time (aspect ratio of face and eye position) are determined, and the degree of conformity to the antecedent part of rule A is obtained.

Then, in step S360, the address pointer i is reset to [0], and in step S362 it is determined whether the address pointer i is equal to the end address. If it is not the end address, the data (consequent part data) E is read from the YY table of the table ROM 12 in which the membership function "age is young" is stored with i as the address in the following step S364. The data in this case corresponds to the goodness of fit of the membership function YY shown in FIG. 2C, and has a value in the range of 0 to 1. The data E is stored in the register of the same name in the CPU 1, for example.

Next, in step S366, the data D and the data E are compared, and if D> E, the process proceeds to step S368 and P stored in the work RAM 9 with i as the address.
Write data E to table A. On the other hand, when D ≦ E, the process proceeds to step S370, and the data D is written in the PA table stored in the work RAM 9 using i as an address. In this way, the processing of cutting the membership function YY is performed by using the smaller value. This processing is to perform a "so-called head-cutting method" in which the membership function YY which is the consequent part is cut according to the conformity of the antecedent part.

Next, in step S372, the address pointer i is incremented by [1], and then the process returns to step S362. Then, in step S362, the same process is repeated until the address pointer i becomes equal to the end address, and when the address pointer i becomes equal to the end address, the process returns to the age estimation switch interrupt routine of FIG. By incrementing the address pointer i to the end address, all membership functions YY will be searched. This allows
The "head cut method" is executed for the membership function YY according to the suitability of the antecedent part. In this way, the fuzzy rule A is calculated on the basis of the aspect ratio of the face and the position of the eyes, and the process is “young when the aspect ratio of the face is large and the position of the eyes is low”. The degree of conformity is calculated. In other words, the fuzzy estimate that the age is young is the truncated membership function YY
Is calculated as the size of.

FIG. 12 is a flow chart showing a subroutine of fuzzy rule B calculation processing. First, in step S400, data (antecedent data) Da from the FM table in the table ROM 12 in which the membership function "medium aspect ratio" is stored, using the contents of the A register (face aspect ratio) as an address. Read out. The data in this case corresponds to the goodness of fit (grade) of the membership function FM shown in FIG. 2A, and has a value in the range of 0 to 1. That is, the data Da is the matching degree indicated by the rule B with respect to the input value of the aspect ratio of the face. Data D
a is stored in a register of the same name in the CPU 1, for example,
This also applies to data Db described later.

Similarly, in step S402, the contents (eye position) of the B register are used as addresses, and the table ROM 1
The data (antecedent part data) Db is read from the EM table in which the membership function “the position of the eyes is medium” in 2 is stored. The data in this case is shown in Figure 2.
It corresponds to the goodness of fit of the membership function EM shown in (b) and is a value in the range of 0 to 1. That is, the data Db is the matching degree indicated by the rule A for the input value of the eye position. Next, in step S404, it is determined whether or not the data D is greater than or equal to the data Db. If Da ≧ Db, the process proceeds to step S406, and the smaller value is adopted as the data D. In this case, D = Db.

This process is to take the logical product of the respective data and is equivalent to the so-called "take MIN". This is to obtain a state that satisfies at least both of two input values by "taking MIN". As a result, the smallest one of the data Da and the data Db is adopted as the determination indicated by the rule B in this routine. On the other hand, when Da <Db, the process proceeds to step S408, and D = Da is adopted. in this way,
For rule B, two input values at that time (aspect ratio of face and eye position) are determined, and the degree of conformity with the antecedent part of rule B is obtained.

Then, in step S410, the address pointer i is reset to [0], and in step S412 it is determined whether the address pointer i is equal to the end address. If it is not the end address, then in step S414 the YM in which the membership function "age is medium" in the table ROM 12 is stored with i as the address.
Data (consequent part data) E is read from the table. The data in this case corresponds to the goodness of fit of the membership function YM shown in FIG. 2C, and has a value in the range of 0 to 1. The data E is stored in the register of the same name in the CPU 1, for example.

Next, in step S416, the data D and the data E are compared, and if D> E, the process proceeds to step S418, in which P is stored in the work RAM 9 with i as the address.
Write data E to table B. On the other hand, when D ≦ E, the process proceeds to step S420, and the data D is written in the PB table stored in the work RAM 9 using i as an address. In this way, the processing of cutting the membership function YM is performed by using the smaller value. This processing is a "so-called head-cutting method" in which the membership function YM, which is the consequent part, is cut in accordance with the degree of conformity of the antecedent part.

Next, in step S422, the address pointer i is incremented by [1], and then the process returns to step S412. Then, in step S412, the same processing is repeated until the address pointer i becomes equal to the end address, and when the address pointer i becomes equal to the end address, the process returns to the age estimation switch interrupt routine of FIG. By incrementing the address pointer i to the end address, all membership functions YM will be searched. This allows
The "head cut method" according to the conformity of the antecedent part is executed for the membership function YM. In this manner, the fuzzy rule B is calculated based on the aspect ratio of the face and the position of the eyes, and "if the aspect ratio of the face is medium and the position of the eyes is medium, the age is middle age. The degree of conformity to the process of “” is calculated. That is, the fuzzy estimated value that the age is middle age is obtained as the magnitude of the truncated membership function YM.

FIG. 13 is a flow chart showing a subroutine of the fuzzy rule C arithmetic processing. First, in step S450, data (antecedent data) Da is obtained from the FS table in the table ROM 12 in which the membership function "aspect ratio is small" is stored with the contents of the A register (face aspect ratio) as an address. read out. The data in this case corresponds to the goodness of fit (grade) of the membership function FS shown in FIG. 2A, and has a value in the range of 0 to 1. That is, the data Da is the matching degree indicated by the rule C with respect to the input value of the aspect ratio of the face. Data D
a is stored in a register of the same name in the CPU 1, for example,
This also applies to data Db described later.

Similarly, in step S452, the contents of the B register (eye position) are used as addresses, and the table ROM1
The data (antecedent part data) Db is read from the EH table in which the membership function "the eye position is high" in 2 is stored. The data in this case is shown in Figure 2.
It corresponds to the goodness of fit of the membership function EH shown in (b) and is a value in the range of 0 to 1. That is, the data Db is the matching degree indicated by the rule C for the input value of the eye position. Next, in step S454, it is determined whether or not the data Da is greater than or equal to the data Db. If Da ≧ Db, the process proceeds to step S456, and the smaller value is adopted as the data D. In this case, D = Db.

This processing is to take the logical product of the respective data and is equivalent to the so-called "take MIN". This is to obtain a state that satisfies at least both of two input values by "taking MIN". As a result, the smallest one of the data Da and the data Db is adopted as the determination indicated by the rule C in this routine. On the other hand, when Da <Db, the process proceeds to step S458 and D = Da is adopted. in this way,
For the rule C, two input values at that time (aspect ratio of the face and eye position) are determined, and the degree of conformity with the antecedent part of the rule C is obtained.

Next, in step S460, the address pointer i is reset to [0], and in step S462 it is determined whether the address pointer i is equal to the end address. If it is not the end address, the data (consequent part data) E is read out from the YO table of the table ROM 12 in which the membership function "Age is high" is stored with the address i in step S464. The data in this case corresponds to the goodness of fit of the membership function YO shown in FIG. 2C, and has a value in the range of 0 to 1. The data E is stored in the register of the same name in the CPU 1, for example.

Next, in step S466, the data D and the data E are compared, and if D> E, the process proceeds to step S468 and P stored in the work RAM 9 with i as the address.
Write the data E in the C table. On the other hand, when D ≦ E, the process proceeds to step S470 and similarly, the data D is written in the PC table stored in the work RAM 9 using i as an address. In this way, the processing of cutting the membership function YO is performed by using the smaller value. This process is a "so-called head-cutting method" in which the membership function YO, which is the consequent part, is cut in accordance with the degree of conformity of the antecedent part.

Then, in step S472, the address pointer i is incremented by [1], and then the process returns to step S462. Then, in step S462, the same process is repeated until the address pointer i becomes equal to the end address, and when the address pointer i becomes equal to the end address, the process returns to the age estimation switch interrupt routine of FIG. By incrementing the address pointer i to the end address, all membership functions YO will be searched. This allows
The "head cut method" is executed for the membership function YO according to the suitability of the antecedent part. In this way, the fuzzy rule C is calculated on the basis of the aspect ratio of the face and the position of the eyes, and in contrast to the process "when the aspect ratio of the face is small and the eye position is high, the age is high". The degree of conformity is calculated. In other words, the fuzzy estimate that the age is high is the truncated membership function YO
Is calculated as the size of.

FIG. 14 is a flowchart showing a subroutine of maximum value calculation processing. First, in step S500, the address pointer i is reset to [0], and step S
Work R with address pointer i as an address in 502
The data D1 is read from the PA memory in the AM9 and loaded into the register of the same name. The data D1 is the calculation result for the fuzzy rule A, and specifically,
It corresponds to the membership function YY, which is calculated by calculating the degree of conformity with the process “Your age is high when the aspect ratio of the face is large and the eye position is low”.

Next, in step S504, the P stored in the work RAM 9 is set with the address pointer i as an address.
The data D2 is read from the B memory and loaded into the register of the same name. The data D2 is the calculation result of the fuzzy rule B, and specifically, the process is “when the aspect ratio of the face is medium and the position of the eyes is medium, the age is middle age”. Corresponding to the membership function YM, which is calculated by calculating the degree of conformity with the membership function. Next, in step S506, the data D3 is similarly read from the PC memory in the work RAM 9 using the address pointer i as an address and loaded into the register of the same name. The data D3 is the calculation result for the fuzzy rule C, and specifically, to what extent is it suitable for the process of “when the aspect ratio of the face is small and the eye position is high, the age is high”? Corresponds to the membership function YO, which is calculated and truncated.

Then, in step S508, the maximum value of the data D1 to D3 is determined. Then, according to the maximum value determination result of the data D1 to D3, the process proceeds to step S510, step S512, and step S514, respectively. Such maximum value determination is performed because it is necessary to take a logical sum (OR logic) to integrate the inference results for each of the fuzzy rules A to C.
This is equivalent to the process "take AX". Specifically, for example, which face image from a fuzzy rule A to a fuzzy rule C is affected by a certain face image, and if it is affected even a little, it responds to a request to reflect it in the judgment of age estimation. This is the basic idea of the "OR logic" relationship. Then, this idea is set as "take a sum" in the set theory, and in this embodiment,
As long as any one is affected, the above process is executed as a target of consideration.

In step S510, the address pointer i is used as an address in the P stored in the work RAM 9.
The contents of the register D1 (data D1) are written in the D memory. As a result, part of the truncated membership function YY, which is the calculation result of the fuzzy rule A specified by the value of the address pointer i (i = 0 in the first routine) this time, is written to the PD memory. As will be described later, by incrementing the address pointer i to the end address, all the truncated membership functions YY are searched and written in the PD memory.

In step S512, the address pointer i
Is used as an address to write the contents of the register D2 (data D2) to the PD memory in the work RAM 9. As a result, part of the truncated membership function YM, which is the calculation result of the fuzzy rule B designated by the value of the address pointer i this time, is written in the PD memory.

Similarly, in step S514, the content of the register D3 (data D3) is written to the PD memory in the work RAM 9 using the address pointer i as an address. As a result, a part of the truncated membership function YO, which is the calculation result of the fuzzy rule C designated by the value of the address pointer i this time, is written in the PD memory. Next, in step S516, it is determined whether the address pointer i is equal to the end address. If it is not the end address, the address pointer i is incremented by [1] in the following step S518, and then the process returns to step S502. Then, in step S516, the above processing is repeated until the address pointer i becomes equal to the end address, and when the address pointer i becomes equal to the end address, the process returns to the age estimation switch interrupt routine of FIG.

By incrementing the address pointer i to the end address, all the truncated membership functions YY, YM, and YO are searched and PD
The data D4 is written in the memory. In this way, the processing of taking a logical sum (“taking MAX”) is performed to integrate the inference results of each of the fuzzy rules A to C, and the truncated membership functions YY, YM, Y
O is OR-combined. That is, the MAX combining processing superimposes the inference results for each of the fuzzy rules A to C to generate a combined output.

FIG. 15 is a flow chart showing a subroutine of the center of gravity calculation processing. First, in step S550, the address pointer i is reset to [0], and the area register area and the center of gravity register haf are similarly reset to [0]. The area register area stores the data of the area area required for the center of gravity calculation, and the center of gravity register haf stores the center of gravity value haf required for the center of gravity calculation. Next, in step S552, the data D4 is read from the PD memory in the work RAM 9 using the address pointer i as an address. Data D4
Is a part of the OR function of the membership functions YY, YM, and YO, which are truncated by the maximum value calculation process described above (the part specified by i as an address). Rule A ~
The inference result for each C is a part of the combined output data (later expressed as an OR combination function).

This time, i (i = 0 in the first routine)
That is, a part of the combined output data is read out from the PD memory by using as the address. As will be described later, by incrementing the address pointer i to the end address, all the combined output data are searched and read. That is, in the following step S554, the value of the area register area is incremented by the combined output data D4 read this time to calculate the integrated value of the area area. Then, in step S556, it is determined whether the address pointer i is equal to the end address. If it is not the end address, the address pointer i is incremented by [1] in the following step S558, and thereafter,
It returns to step S552. Then, in step S556, until the address pointer i becomes equal to the end address,
The above processing is repeated, and when the address pointer i becomes equal to the end address, the procedure goes to step S560.

By incrementing the address pointer i to the end address, all OR-combined output data of the truncated membership functions YY, YM and YO are searched and read. Step S
At 560, the integrated area are obtained at step S554
The value of 1/2 of a is set as the center of gravity value haf and the center of gravity register ha
Store in f. Next, in step S562, both j and the area integration area bal corresponding to the horizontal axis of the OR combination function of the membership functions YY, YM, and YO are reset to [0]. The values of j and bal are stored in the register having the same name, for example.

Next, in step S564, the data D4 corresponding to the OR combination function is read from the PD memory in the work RAM 9 using the horizontal axis j as an address. In the following step S566, the value of the area integration area bal is incremented by the data D4 read this time to integrate the area integration area bal. Next, the area integration area bal integrated in step S568 is compared with the center of gravity value haf. When the area integration area bal is smaller than the center of gravity value haf, the process proceeds to step S570 to increment the horizontal axis j and then returns to step S564. And step S
At 568, the above processing is repeated until the area integration area bal becomes equal to or more than the center of gravity value haf, and when bal ≧ haf,
It is determined that the value of the horizontal axis j at that time is the center of gravity, the value of the horizontal axis i is displayed as the estimated age on the display unit 13 in step S572, and the process returns to the age estimation switch interrupt routine of FIG. At the same time, the estimated age value is stored in the register in the CPU 1. In this way, O
The defuzzification process (defuzzifier process) that takes the center of gravity of the R composition function and uses that value as a representative of the logical sum is performed, and the entire conclusion in the inference process is one definite value, that is, the age in this case. Is calculated as an estimated value of.

As described above, in this embodiment, when it is desired to know the age of the face image created by operating the cursor switch 3, the position correction switch 5, etc., the age estimation switch 6 is pushed to create the created face. Characteristic data (here, the aspect ratio of the face and the position of the eyes) according to the age of the image are detected, fuzzy inference is performed according to the fuzzy rules A to C described above using the detected data as input parameters, and the inference result is obtained. The age of the face image created based on this is estimated and calculated as the value of the horizontal axis j (center of gravity point) by the routine of the center of gravity calculation processing of FIG. The calculated estimated age is, as shown in FIG.
It is displayed on the display unit 13 together with F4.

Therefore, the age can be objectively and quickly estimated from the created face image without requiring special skill, and the result of comparison with the average age can be easily known. You can also And age input key 7
Since the actual input age is additionally displayed as shown in FIG. 16, it is possible to compare the actual age and the estimated age estimated based on the face image in comparison with the actual age. Therefore, it is possible to judge whether it looks younger than the actual age or looks older. In addition, since fuzzy inference is used in the age estimation process, a huge amount of processing is not required in the estimation calculation, and the calculation load can be reduced. As a result, the existing CPU can easily perform the calculation, and the highly functional age estimating device can be realized at low cost.

In the above-described embodiment, the cursor switch 3 is operated to select a part pattern for each part stored in the parts screen ROM 10, or the position of the selected part pattern is adjusted using the position correction switch 5. Although the age is estimated by fuzzy inference based on the face image created by correcting, for example, the face image data representing the face image captured by the CCD camera 20 is stored in the face image RAM 21. The CPU 1 detects the aspect ratio of the face, the position of the eyes, and the like based on the stored face image data, and executes fuzzy inference based on these face feature data to estimate its age. Good. Further, in the above-mentioned embodiment, when estimating the age of the face image created in advance by fuzzy inference, the face feature data representing the aspect ratio of the face and the eye position is used as the input data, but as the face feature data, In addition to this, the number of wrinkles on the face, the degree of receding hair (the degree of baldness of the forehead), and the like may be used. Further, in the above embodiment, the fuzzy inference is realized by software using the CPU 1 and the table ROM 12, but it may be realized by hardware using a fuzzy chip, for example. Also, the value of the membership function used in fuzzy inference is the target image,
It may be set appropriately according to the characteristic data. By doing so, the accuracy of age estimation is further improved.

By dividing by sex,
With respect to the face of a woman, a face image according to age can be automatically created by the same process, and the age can be objectively and quickly estimated from the created face image.
Further, the face image may be one's own face or the face of another person. For example, if it is a face image of another person, create it as a montage image, and when registering it with its personal data (at this time, for example, excluding age data), estimate its age and register it as personal data. You can also In addition, a face image to be registered in advance is created using this apparatus, and it is stored in another storage means (for example, EE).
It may be stored in a PROM or a non-volatile RAM). By doing so, even if the power is turned off, the latest registered face image is used to create a past or future expected face, and if the age is estimated with respect to the past or future expected face, the average age will be obtained. It is also possible to easily know the comparison result of.

[0087]

According to the first aspect of the invention, a face image is created by combining a plurality of part images, the features of the created face image are detected, and then the detected features are used as input parameters. Fuzzy inference can be performed according to a predetermined fuzzy rule, and the age of the created face image can be estimated based on the result of the fuzzy inference. Not only that, but the actual age entered and displayed
And the estimated age estimated based on the facial image
Compare and examine while viewing the face or face image being displayed.
And the displayed face image looks younger than the actual age.
Quickly and reliably confirm whether or not you look old
be able to. According to the second aspect of the present invention, a facial feature is detected, fuzzy inference is performed according to a predetermined fuzzy rule using the facial feature data representing the detected facial feature as input data, and the result of the fuzzy inference is obtained. Based on this, the age of the face can be estimated. Not only that, but based on the actual age entered and displayed and the face
The estimated age and estimated size by the face or face being displayed.
You can compare and examine while looking at the image, and the face displayed is
Judging whether you look younger or older than the actual age
Can be confirmed quickly and reliably. Further, according to the invention of claim 3, to display a face it is stored, and detects the facial features being the display, based on the characteristics of the detected face, to estimate the age of the face be able to. Not only that, but the actual typed and displayed
The size of the age and the estimated age estimated based on the face
Compare and examine while viewing the face or face image being displayed.
The displayed face looks younger than the actual age, or
Quick and reliable confirmation of what looks old
it can. Therefore, according to the present invention, it is possible not only to quickly and easily estimate an objective age from a face or a face image without requiring special skills , but also to input
Based on the actual age displayed and the face or face image.
The estimated age and
You can compare and examine while looking at the face image and display the face
Or the face image looks younger than the actual age or looks old.
It is possible to quickly and surely confirm whether or not to make a decision.

[Brief description of drawings]

FIG. 1 is a configuration diagram of an embodiment of an age estimating device according to the present invention.

FIG. 2 is a diagram showing a fuzzy inference membership function of the embodiment.

FIG. 3 is a flowchart showing a main program of face image creation / age estimation processing of the embodiment.

FIG. 4 is a flowchart showing a feature switch interrupt routine of the same embodiment.

FIG. 5 is a diagram showing an example of a data storage area of the work RAM of the same embodiment.

FIG. 6 is a diagram showing an example of screen data for each part stored in a parts screen ROM of the embodiment.

FIG. 7 is a flowchart showing a cursor switch interrupt routine of the embodiment.

FIG. 8 is a flowchart showing a display switch interrupt routine of the embodiment.

FIG. 9 is a flowchart showing a position correction switch interrupt routine of the embodiment.

FIG. 10 is a flowchart showing an age estimation switch interrupt routine of the embodiment.

FIG. 11 is a flowchart showing a subroutine of fuzzy rule A calculation processing of the embodiment.

FIG. 12 is a flowchart showing a subroutine of fuzzy rule B calculation processing of the embodiment.

FIG. 13 is a flowchart showing a subroutine of fuzzy rule C arithmetic processing of the embodiment.

FIG. 14 is a flowchart showing a subroutine of maximum value calculation processing of the embodiment.

FIG. 15 is a flowchart showing a subroutine of a center of gravity calculation process of the same embodiment.

FIG. 16 is a diagram showing an example of a created face image of the embodiment.

[Explanation of symbols]

1 CPU (face image creation means, feature data detection means, face feature detection means, actual age designation setting means, age estimation means) 2 feature switch 3 cursor switch 4 display switch 5 position correction switch 6 age estimation switch 7 age input key 8 Program ROM 9 Work RAM 10 Parts screen ROM (parts pattern storage means) 11 VRAM 12 Table ROM (face characteristic data storage means) 13 Display unit (display means)

Continuation of the front page (56) References JP-A-6-168314 (JP, A) Kuniaki Iida Tatsuya Furukawa, Age fuzzy reasoning of growing person based on frontal skeleton data, IEICE Transactions J75 -D-2, Japan, The Institute of Electronics, Information and Communication Engineers, March 1992, p538-544 (58) Fields investigated (Int.Cl. ⁷ , DB name) G06T 1/00 G06T 7/00-7 / 60 G06T 11/00-11/80 A63F 13/00

Claims (9)

(57) [Claims] 1. A face image is formed by combining a plurality of part images.
Face image creation means to create, The features of the face image created by this face image creating means
Facial feature detection means for detecting, Input the features detected by this facial feature detection means.
Fuzzy inference according to predetermined fuzzy rules as data
Based on the result of this fuzzy inference
Estimate the age of the face image created by the image creation means
Age estimation means,It corresponds to the face image created by the face image creating means.
Age input means to input the actual age, The actual age input by this age input means, the year
The estimated age estimated by the age estimating means, and
The face image created by the face image creating means is displayed on the same screen.
Display control means for controlling to display An age estimating device comprising: 2. A facial feature detecting means for detecting facial features, A face representing the features of the face detected by the face feature detecting means.
Use the characteristic data as input data and follow a predetermined fuzzy rule
Therefore, fuzzy inference is performed, and the result of this fuzzy inference is
Age estimating means for estimating the age of the face based onActual corresponding to the face detected by the face feature detection means
Age input means to enter the age of, The actual age input by this age input means, the year
The estimated age estimated by the age estimating means, and
The face detected by the face feature detection means is displayed on the same screen.
Display control means for controlling to display, An age estimating device comprising: 3. A first table for displaying stored faces.
Indicating control means, The face displayed by the control of the first display control means.
Facial feature detecting means for detecting the feature of Based on the facial features detected by this facial feature detection means
And an age estimating means for estimating the age of the face,After displaying the face under the control of the first display control means,
Age input means for inputting the actual age corresponding to the face
When, Actual age and the year entered by this age input means
The estimated age estimated by the age estimation means is displayed above.
Second control to display on the same screen as the face
Display control means of An age estimating device comprising: 4. The face characteristic is any one of a face aspect ratio, eye height with respect to the whole face, eye size, number of wrinkles, and receding degree of hair. Item 2 or 3
The described age estimation device. 5. The face feature detecting means is at least one of aspect ratio of the face image created by the face image creating means, eye height, eye size, number of wrinkles, and receding degree of hair. The age estimating device according to claim 1, wherein one is detected. 6. The age estimating means estimates the age of the face by performing fuzzy inference according to the aspect ratio of the face and the eye height which are the features of the face.
The described age estimation device. 7.Combine multiple parts images to create a face image
Face image creation step to create, ThisOf the face image created by the face image creation step of
A facial feature detecting step for detecting a sign, Input the features detected by this facial feature detection step.
Fuzzy according to a predetermined fuzzy rule as a parameter
Make inferences, and based on the results of this fuzzy inference,
The age of the face image created by the face image creation step
An age estimation step to estimate, Corresponds to the face image created by the face image creation step
An age input step to enter the actual age, Actual age entered in this age entry step, previous
Estimated age estimated by the age estimation step, and
And the face image created in the face image creating step
Display control step that controls to display on the same screen.
And Age control method characterized by comprising . 8.Facial feature detection step for detecting facial features
When, Display the facial features detected by this facial feature detection step.
Using the facial feature data as input data,
Fuzzy inference according to
Age estimation step for estimating the age of the face based on the result
When, Corresponds to the face detected by the face feature detection step
An age input step to enter the actual age, Actual age entered in this age entry step, previous
Estimated age estimated by the age estimation step, and
The face detected by the face feature detection step
Display control step to control to display on the same screen
When, An age display control method comprising: 9.First table for displaying stored faces
The indicated control step, It is displayed under the control of this first display control step.
Facial feature detection step for detecting facial features that Based on the facial features detected by this facial feature detection step,
Then, an age estimation step of estimating the age of the face, The face is displayed under the control of the first display control step.
Then enter the actual age corresponding to the face
Tep, Actual age and previous entered by this age input step
The estimated age estimated by the age estimation step
Control to display on the same screen as the displayed face
A second display control step to An age display control method comprising: